Fix Prometheus queries after adding Per Orchestrator metrics

[leszko]

In livepeer/go-livepeer#2371 we tagged 3 metrics with the orchestrator address and service URI labels. This change may not be backwards-compatible for some queries we use. In particular, it's an issue for this query we use to get average round trip time:

rate(livepeer_transcode_overall_latency_seconds_sum[1m])/rate(livepeer_transcode_overall_latency_seconds_count[1m])

Since we get always the first entry in from the query, it resulted in the bug that not every stream has latency equal to zero.